def __call__(self, prefix: str):
logging.info(f'aggregate skeletons with prefix of {prefix}')
id2filenames = defaultdict(list)
for filename in self.fragments_storage.list_files(prefix=prefix):
filename = os.path.basename(filename)
# `match` implies the beginning (^). `search` matches whole string
matches = re.search(r'(\d+):', filename)
if not matches:
continue
# skeleton ID
skl_id = int(matches.group(0)[:-1])
id2filenames[skl_id].append(filename)
for skl_id, filenames in id2filenames.items():
logging.info(f'skeleton id: {skl_id}')
frags = self.fragments_storage.get(filenames)
frags = [
PrecomputedSkeleton.from_precomputed(x['content'])
for x in frags
]
skel = PrecomputedSkeleton.simple_merge(frags).consolidate()
skel = kimimaro.postprocess(skel,
dust_threshold=1000,
tick_threshold=3500)
self.output_storage.put(
file_path=str(skl_id),
content=skel.to_precomputed(),
)
# the last few hundred files will not be uploaded without sleeping!
sleep(0.01)
def test_caching():
vol = CloudVolume('file:///tmp/cloudvolume/test-skeletons',
info=info,
cache=True)
vol.cache.flush()
skel = PrecomputedSkeleton(
[
(0, 0, 0),
(1, 0, 0),
(2, 0, 0),
(0, 1, 0),
(0, 2, 0),
(0, 3, 0),
],
edges=[(0, 1), (1, 2), (3, 4), (4, 5), (3, 5)],
segid=666,
)
vol.skeleton.upload(skel)
assert vol.cache.list_skeletons() == ['666.gz']
skel.id = 1
with open(os.path.join(vol.cache.path, 'skeletons/1'), 'wb') as f:
f.write(skel.encode())
cached_skel = vol.skeleton.get(1)
assert cached_skel == skel
vol.cache.flush()
def test_consolidate():
skel = PrecomputedSkeleton(
vertices=np.array([
(0, 0, 0),
(1, 0, 0),
(2, 0, 0),
(0, 0, 0),
(2, 1, 0),
(2, 2, 0),
(2, 2, 1),
(2, 2, 2),
],
dtype=np.float32),
edges=np.array([
[0, 1],
[1, 2],
[2, 3],
[3, 4],
[4, 5],
[5, 6],
[6, 7],
],
dtype=np.uint32),
radii=np.array([0, 1, 2, 3, 4, 5, 6, 7], dtype=np.float32),
vertex_types=np.array([0, 1, 2, 3, 4, 5, 6, 7], dtype=np.uint8),
)
correct_skel = PrecomputedSkeleton(
vertices=np.array([
(0, 0, 0),
(1, 0, 0),
(2, 0, 0),
(2, 1, 0),
(2, 2, 0),
(2, 2, 1),
(2, 2, 2),
],
dtype=np.float32),
edges=np.array([
[0, 1],
[0, 2],
[0, 3],
[1, 2],
[3, 4],
[4, 5],
[5, 6],
],
dtype=np.uint32),
radii=np.array([0, 1, 2, 4, 5, 6, 7], dtype=np.float32),
vertex_types=np.array([0, 1, 2, 4, 5, 6, 7], dtype=np.uint8),
)
consolidated = skel.consolidate()
assert np.all(consolidated.vertices == correct_skel.vertices)
assert np.all(consolidated.edges == correct_skel.edges)
assert np.all(consolidated.radii == correct_skel.radii)
assert np.all(consolidated.vertex_types == correct_skel.vertex_types)
def test_cable_length():
skel = PrecomputedSkeleton([
(0,0,0), (1,0,0), (2,0,0), (3,0,0), (4,0,0), (5,0,0)
],
edges=[ (1,0), (1,2), (2,3), (3,4), (5,4) ],
radii=[ 1, 2, 3, 4, 5, 6 ],
vertex_types=[1, 2, 3, 4, 5, 6]
)
assert skel.cable_length() == (skel.vertices.shape[0] - 1)
skel = PrecomputedSkeleton([
(2,0,0), (1,0,0), (0,0,0), (0,5,0), (0,6,0), (0,7,0)
],
edges=[ (1,0), (1,2), (2,3), (3,4), (5,4) ],
radii=[ 1, 2, 3, 4, 5, 6 ],
vertex_types=[1, 2, 3, 4, 5, 6]
)
assert skel.cable_length() == 9
skel = PrecomputedSkeleton([
(1,1,1), (0,0,0), (1,0,0)
],
edges=[ (1,0), (1,2) ],
radii=[ 1, 2, 3],
vertex_types=[1, 2, 3]
)
assert abs(skel.cable_length() - (math.sqrt(3) + 1)) < 1e-6
def test_read_swc():
# From http://research.mssm.edu/cnic/swc.html
test_file = """# ORIGINAL_SOURCE NeuronStudio 0.8.80
# CREATURE
# REGION
# FIELD/LAYER
# TYPE
# CONTRIBUTOR
# REFERENCE
# RAW
# EXTRAS
# SOMA_AREA
# SHINKAGE_CORRECTION 1.0 1.0 1.0
# VERSION_NUMBER 1.0
# VERSION_DATE 2007-07-24
# SCALE 1.0 1.0 1.0
1 1 14.566132 34.873772 7.857000 0.717830 -1
2 0 16.022520 33.760513 7.047000 0.463378 1
3 5 17.542000 32.604973 6.885001 0.638007 2
4 0 19.163984 32.022469 5.913000 0.602284 3
5 0 20.448090 30.822802 4.860000 0.436025 4
6 6 21.897903 28.881084 3.402000 0.471886 5
7 0 18.461960 30.289471 8.586000 0.447463 3
8 6 19.420759 28.730757 9.558000 0.496217 7"""
skel = PrecomputedSkeleton.from_swc(test_file)
assert skel.vertices.shape[0] == 8
assert skel.edges.shape[0] == 7
skel_gt = PrecomputedSkeleton(
vertices=[
[14.566132, 34.873772, 7.857000],
[16.022520, 33.760513, 7.047000],
[17.542000, 32.604973, 6.885001],
[19.163984, 32.022469, 5.913000],
[20.448090, 30.822802, 4.860000],
[21.897903, 28.881084, 3.402000],
[18.461960, 30.289471, 8.586000],
[19.420759, 28.730757, 9.558000]
],
edges=[ (0,1), (1,2), (2,3), (3,4), (4,5), (2,6), (7,6) ],
radii=[
0.717830, 0.463378, 0.638007, 0.602284,
0.436025, 0.471886, 0.447463, 0.496217
],
vertex_types=[
1, 0, 5, 0, 0, 6, 0, 6
],
)
assert PrecomputedSkeleton.equivalent(skel, skel_gt)
def test_downsample_joints():
skel = PrecomputedSkeleton([
(2, 3,0), # 0
(2, 2,0), # 1
(2, 1,0), # 2
(0,0,0), (1,0,0), (2, 0,0), (3,0,0), (4,0,0), # 3, 4, 5, 6, 7
(2,-1,0), # 8
(2,-2,0), # 9
(2,-3,0), # 10
],
edges=[
(0, 1),
(1, 2),
(2, 5),
(3,4), (4,5), (5, 6), (6,7),
(5, 8),
(8, 9),
(9,10)
],
radii=[ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 ],
vertex_types=[ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 ],
segid=1337,
)
ds_skel = skel.downsample(2)
ds_skel_gt = PrecomputedSkeleton([
(2, 3,0), # 0
(2, 2,0), # 1
(0,0,0), (2, 0,0), (4,0,0), # 2, 3, 4
(2,-2,0), # 5
(2,-3,0), # 6
],
edges=[
(0,1),
(1,3),
(2,3), (3,4),
(3,5),
(5,6)
],
radii=[ 0, 1, 3, 5, 7, 9, 10 ],
vertex_types=[ 0, 1, 3, 5, 7, 9, 10 ],
segid=1337,
)
assert PrecomputedSkeleton.equivalent(ds_skel, ds_skel_gt)
def fuse_skeletons(self, skels):
if len(skels) == 0:
return PrecomputedSkeleton()
bbxs = [ item[0] for item in skels ]
skeletons = [ item[1] for item in skels ]
skeletons = self.crop_skels(bbxs, skeletons)
skeletons = [ s for s in skeletons if not s.empty() ]
if len(skeletons) == 0:
return PrecomputedSkeleton()
return PrecomputedSkeleton.simple_merge(skeletons).consolidate()
def process_skeletons(self, unfused_skeletons, in_place=False):
skeletons = {}
if in_place:
skeletons = unfused_skeletons
for label in tqdm(unfused_skeletons.keys(),
desc="Postprocessing",
disable=(not self.progress)):
skels = unfused_skeletons[label]
skel = PrecomputedSkeleton.simple_merge(skels)
skel.id = label
skel.extra_attributes = [
attr for attr in skel.extra_attributes \
if attr['data_type'] == 'float32'
]
if self.max_cable_length is not None and skel.cable_length(
) > self.max_cable_length:
skeletons[label] = skel.to_precomputed()
else:
skeletons[label] = kimimaro.postprocess(
skel,
dust_threshold=self.dust_threshold, # voxels
tick_threshold=self.tick_threshold, # nm
).to_precomputed()
return skeletons
def remove_ticks(skeleton, threshold):
"""
Simple merging of individual TESAR cubes results in lots of little
ticks due to the edge effect. We can remove them by thresholding
the path length from a given branch to the "main body" of the neurite.
We successively remove paths from shortest to longest until no branches
below threshold remain.
If TEASAR parameters were chosen such that they allowed for spines to
be traced, this is also an opportunity to correct for that.
This algorithm is O(N^2) in the number of terminal nodes.
Parameters:
threshold: The maximum length in nanometers that may be culled.
Returns: tick free skeleton
"""
if skeleton.empty():
return skeleton
skels = []
for component in skeleton.components():
skels.append(_remove_ticks(component, threshold))
return PrecomputedSkeleton.simple_merge(skels).consolidate()
def readSkelFromFile(fname, iz, iy, ix, bsize, anisotropy):
print("Reading from file: " + fname)
with open(fname, 'r') as f:
inp_swc = f.read()
skel_read = PrecomputedSkeleton.from_swc(inp_swc)
# revert order from x y z to z y x
vertices_copy = skel_read.vertices.copy()
skel_read.vertices[:, 0] = vertices_copy[:, 2]
skel_read.vertices[:, 2] = vertices_copy[:, 0]
x_offset = ix * bsize[OR_X] * anisotropy[OR_X]
y_offset = iy * bsize[OR_Y] * anisotropy[OR_Y]
z_offset = iz * bsize[OR_Z] * anisotropy[OR_Z]
transform_ = np.array([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0]])
transform_[0, 3] = z_offset
transform_[1, 3] = y_offset
transform_[2, 3] = x_offset
skel_read.transform = transform_
skel_read.apply_transform()
skel_read.transform = np.array([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0]])
return skel_read
def execute(self):
corgie_logger.info(f"Merging skeletons at {self.dst_path}")
fragment_filenames = self.cf.list(prefix=self.prefix, flat=True)
skeleton_files = self.cf.get(fragment_filenames)
skeletons = defaultdict(list)
for skeleton_file in skeleton_files:
try:
colon_index = skeleton_file["path"].index(":")
except ValueError:
# File is full skeleton, not fragment
continue
seg_id = skeleton_file["path"][0:colon_index]
skeleton_fragment = pickle.loads(skeleton_file["content"])
if not skeleton_fragment.empty():
skeletons[seg_id].append(skeleton_fragment)
for seg_id, skeleton_fragments in skeletons.items():
skeleton = PrecomputedSkeleton.simple_merge(
skeleton_fragments).consolidate()
skeleton = kimimaro.postprocess(skeleton, self.dust_threshold,
self.tick_threshold)
skeleton.id = int(seg_id)
self.cf.put(path=seg_id,
content=skeleton.to_precomputed(),
compress="gzip")
corgie_logger.info(f"Finished skeleton {seg_id}")
def merge(skeletons):
merged_skels = {}
for segid, skels in skeletons.items():
skel = PrecomputedSkeleton.simple_merge(skels)
merged_skels[segid] = skel.consolidate()
return merged_skels
def test_components():
skel = PrecomputedSkeleton(
[
(0, 0, 0),
(1, 0, 0),
(2, 0, 0),
(0, 1, 0),
(0, 2, 0),
(0, 3, 0),
],
edges=[(0, 1), (1, 2), (3, 4), (4, 5), (3, 5)],
segid=666,
)
components = skel.components()
assert len(components) == 2
assert components[0].vertices.shape[0] == 3
assert components[1].vertices.shape[0] == 3
assert components[0].edges.shape[0] == 2
assert components[1].edges.shape[0] == 3
skel1_gt = PrecomputedSkeleton([(0, 0, 0), (1, 0, 0), (2, 0, 0)], [(0, 1),
(1, 2)])
skel2_gt = PrecomputedSkeleton([(0, 1, 0), (0, 2, 0), (0, 3, 0)], [(0, 1),
(0, 2),
(1, 2)])
assert PrecomputedSkeleton.equivalent(components[0], skel1_gt)
assert PrecomputedSkeleton.equivalent(components[1], skel2_gt)
def remove_loops(skeleton):
if skeleton.empty():
return skeleton
skels = []
for component in skeleton.components():
skels.append(_remove_loops(component))
return PrecomputedSkeleton.simple_merge(skels).consolidate()
def test_remove_disconnected_vertices():
skel = PrecomputedSkeleton(
[
(0,0,0), (1,0,0), (2,0,0),
(0,1,0), (0,2,0), (0,3,0),
(-1, -1, -1)
],
edges=[
(0,1), (1,2),
(3,4), (4,5), (3,5)
],
segid=666,
)
res = skel.remove_disconnected_vertices()
assert res.vertices.shape[0] == 6
assert res.edges.shape[0] == 5
assert res.radii.shape[0] == 6
assert res.vertex_types.shape[0] == 6
assert res.id == 666
def remove_dust(skeleton, dust_threshold):
"""Dust threshold in physical cable length."""
if skeleton.empty():
return skeleton
skels = []
for skel in skeleton.components():
if skel.cable_length() > dust_threshold:
skels.append(skel)
skeleton = PrecomputedSkeleton.simple_merge(skels)
return skeleton.consolidate()
def process_skeletons(self, locations, cv):
skeletons = {}
for label, locs in locations.items():
skel = PrecomputedSkeleton.simple_merge(
self.get_unfused(label, locs, cv)
)
skel.id = label
skel.extra_attributes = [
attr for attr in skel.extra_attributes \
if attr['data_type'] == 'float32'
]
skeletons[label] = kimimaro.postprocess(
skel,
dust_threshold=self.dust_threshold, # voxels
tick_threshold=self.tick_threshold, # nm
).to_precomputed()
return skeletons
def load_raw_skeletons():
print("Downloading list of files...")
print(cv.skeleton.meta.layerpath)
with Storage(cv.skeleton.meta.layerpath, progress=True) as stor:
all_files = list(stor.list_files())
all_files = [
fname for fname in all_files if os.path.splitext(fname)[1] == '.frags'
]
print("Downloading files...")
with Storage(cv.skeleton.meta.layerpath, progress=True) as stor:
all_files = stor.get_files(all_files)
# CHECKPOINT?
for i, res in enumerate(tqdm(all_files, desc='Unpickling')):
all_files[i] = pickle.loads(res['content'])
# group by segid
unfused_skeletons = defaultdict(list)
while all_files:
fragment = all_files.pop()
for label, skel_frag in fragment.items():
unfused_skeletons[label].append(skel_frag)
# CHECKPOINT?
skeletons = crt_dict()
labels = list(unfused_skeletons.keys())
for label in tqdm(labels, desc='Simple Merging'):
skels = unfused_skeletons[label]
skeleton = PrecomputedSkeleton.simple_merge(skels)
skeleton.id = label
skeleton.extra_attributes = [
attr for attr in skeleton.extra_attributes \
if attr['data_type'] == 'float32'
]
skeletons[label] = skeleton
del unfused_skeletons[label]
return skeletons
def process_skeletons(self, locations, cv):
filenames = set(itertools.chain(*locations.values()))
labels = set(locations.keys())
unfused_skeletons = self.get_unfused(labels, filenames, cv)
skeletons = {}
for label, skels in tqdm(unfused_skeletons.items(),
desc="Postprocessing",
disable=(not self.progress)):
skel = PrecomputedSkeleton.simple_merge(skels)
skel.id = label
skel.extra_attributes = [
attr for attr in skel.extra_attributes \
if attr['data_type'] == 'float32'
]
skeletons[label] = kimimaro.postprocess(
skel,
dust_threshold=self.dust_threshold, # voxels
tick_threshold=self.tick_threshold, # nm
).to_precomputed()
return skeletons
def readSkelFromFile(fname, bz, by, bx, bsize, anisotropy):
print("Reading from file: " + fname)
with open(fname, 'r') as f:
inp_swc = f.read()
skel_read = PrecomputedSkeleton.from_swc(inp_swc)
x_offset = bx*bsize[0]*anisotropy[0]
y_offset = by*bsize[1]*anisotropy[1]
z_offset = bz*bsize[2]*anisotropy[2]
print("Block: " + str((bx,by,bz)))
print("Offset: "+ str((x_offset,y_offset,z_offset)))
skel_read.transform = np.array([[1,0,0,x_offset],
[0,1,0,y_offset],
[0,0,1,z_offset]])
print("transform: " + str(skel_read.transform))
# print("xmin, xmax: " + str(np.min(skel_read.vertices[:,0])) + ", " + str(np.max(skel_read.vertices[:,0])))
# print("ymin, ymax: " + str(np.min(skel_read.vertices[:,1])) + ", " + str(np.max(skel_read.vertices[:,1])))
# print("zmin, zmax: " + str(np.min(skel_read.vertices[:,2])) + ", " + str(np.max(skel_read.vertices[:,2])))
# print("----------------------------")
#
skel_read.apply_transform()
#
# print("xmin, xmax: " + str(np.min(skel_read.vertices[:,0])) + ", " + str(np.max(skel_read.vertices[:,0])))
# print("ymin, ymax: " + str(np.min(skel_read.vertices[:,1])) + ", " + str(np.max(skel_read.vertices[:,1])))
# print("zmin, zmax: " + str(np.min(skel_read.vertices[:,2])) + ", " + str(np.max(skel_read.vertices[:,2])))
skel_read.transform = np.array([[1,0,0,0],
[0,1,0,0],
[0,0,1,0]])
return skel_read
def trace(
labels,
DBF,
scale=10,
const=10,
anisotropy=(1, 1, 1),
soma_detection_threshold=1100,
soma_acceptance_threshold=4000,
pdrf_scale=5000,
pdrf_exponent=16,
soma_invalidation_scale=0.5,
soma_invalidation_const=0,
fix_branching=True,
manual_targets_before=[],
manual_targets_after=[],
root=None,
max_paths=None,
voxel_graph=None,
):
"""
Given the euclidean distance transform of a label ("Distance to Boundary Function"),
convert it into a skeleton using an algorithm based on TEASAR.
DBF: Result of the euclidean distance transform. Must represent a single label,
assumed to be expressed in chosen physical units (i.e. nm)
scale: during the "rolling ball" invalidation phase, multiply the DBF value by this.
const: during the "rolling ball" invalidation phase, this is the minimum radius in chosen physical units (i.e. nm).
anisotropy: (x,y,z) conversion factor for voxels to chosen physical units (i.e. nm)
soma_detection_threshold: if object has a DBF value larger than this,
root will be placed at largest DBF value and special one time invalidation
will be run over that root location (see soma_invalidation scale)
expressed in chosen physical units (i.e. nm)
pdrf_scale: scale factor in front of dbf, used to weight dbf over euclidean distance (higher to pay more attention to dbf) (default 5000)
pdrf_exponent: exponent in dbf formula on distance from edge, faster if factor of 2 (default 16)
soma_invalidation_scale: the 'scale' factor used in the one time soma root invalidation (default .5)
soma_invalidation_const: the 'const' factor used in the one time soma root invalidation (default 0)
(units in chosen physical units (i.e. nm))
fix_branching: When enabled, zero out the graph edge weights traversed by
of previously found paths. This causes branch points to occur closer to
the actual path divergence. However, there is a large performance penalty
associated with this as dijkstra's algorithm is computed once per a path
rather than once per a skeleton.
manual_targets_before: list of (x,y,z) that correspond to locations that must
have paths drawn to. Used for specifying root and border targets for
merging adjacent chunks out-of-core. Targets are applied before ordinary
target selection.
manual_targets_after: Same as manual_targets_before but the additional
targets are applied after the usual algorithm runs. The current
invalidation status of the shape makes no difference.
max_paths: If a label requires drawing this number of paths or more,
abort and move onto the next label.
root: If you want to force the root to be a particular voxel, you can
specify it here.
voxel_graph: a connection graph that defines permissible
directions of motion between voxels. This is useful for
dealing with self-touches. The graph is defined by the
conventions used in cc3d.voxel_connectivity_graph
(https://github.com/seung-lab/connected-components-3d/blob/3.2.0/cc3d_graphs.hpp#L73-L92)
Based on the algorithm by:
M. Sato, I. Bitter, M. Bender, A. Kaufman, and M. Nakajima.
"TEASAR: tree-structure extraction algorithm for accurate and robust skeletons"
Proc. the Eighth Pacific Conference on Computer Graphics and Applications. Oct. 2000.
doi:10.1109/PCCGA.2000.883951 (https://ieeexplore.ieee.org/document/883951/)
Returns: Skeleton object
"""
dbf_max = np.max(DBF)
labels = np.asfortranarray(labels)
DBF = np.asfortranarray(DBF)
soma_mode = False
# > 5000 nm, gonna be a soma or blood vessel
# For somata: specially handle the root by
# placing it at the approximate center of the soma
if dbf_max > soma_detection_threshold:
labels, num_voxels_filled = fill_voids.fill(labels,
in_place=True,
return_fill_count=True)
if num_voxels_filled > 0:
del DBF
DBF = edt.edt(labels,
anisotropy=anisotropy,
order='F',
black_border=np.all(labels))
dbf_max = np.max(DBF)
soma_mode = dbf_max > soma_acceptance_threshold
soma_radius = 0.0
if soma_mode:
if root is not None:
manual_targets_before.insert(0, root)
root = find_soma_root(DBF, dbf_max)
soma_radius = dbf_max * soma_invalidation_scale + soma_invalidation_const
elif root is None:
root = find_root(labels, anisotropy)
if root is None:
return PrecomputedSkeleton()
free_space_radius = 0 if not soma_mode else DBF[root]
# DBF: Distance to Boundary Field
# DAF: Distance from any voxel Field (distance from root field)
# PDRF: Penalized Distance from Root Field
DBF = kimimaro.skeletontricks.zero2inf(DBF) # DBF[ DBF == 0 ] = np.inf
DAF, target = dijkstra3d.euclidean_distance_field(
labels,
root,
anisotropy=anisotropy,
free_space_radius=free_space_radius,
voxel_graph=voxel_graph,
return_max_location=True,
)
DAF = kimimaro.skeletontricks.inf2zero(DAF) # DAF[ DAF == np.inf ] = 0
PDRF = compute_pdrf(dbf_max, pdrf_scale, pdrf_exponent, DBF, DAF)
# Use dijkstra propogation w/o a target to generate a field of
# pointers from each voxel to its parent. Then we can rapidly
# compute multiple paths by simply hopping pointers using path_from_parents
if not fix_branching:
parents = dijkstra3d.parental_field(PDRF, root, voxel_graph)
del PDRF
else:
parents = PDRF
if soma_mode:
invalidated, labels = kimimaro.skeletontricks.roll_invalidation_ball(
labels,
DBF,
np.array([root], dtype=np.uint32),
scale=soma_invalidation_scale,
const=soma_invalidation_const,
anisotropy=anisotropy)
# This target is only valid if no
# invalidations have occured yet.
elif len(manual_targets_before) == 0:
manual_targets_before.append(target)
# delete reference to DAF and place it in
# a list where we can delete it later and
# free that memory.
DAF = [DAF]
paths = compute_paths(root, labels, DBF, DAF, parents, scale, const,
anisotropy, soma_mode, soma_radius, fix_branching,
manual_targets_before, manual_targets_after,
max_paths, voxel_graph)
skel = PrecomputedSkeleton.simple_merge([
PrecomputedSkeleton.from_path(path) for path in paths if len(path) > 0
]).consolidate()
verts = skel.vertices.flatten().astype(np.uint32)
skel.radii = DBF[verts[::3], verts[1::3], verts[2::3]]
return skel
def test_equivalent():
assert PrecomputedSkeleton.equivalent(PrecomputedSkeleton(),
PrecomputedSkeleton())
identity = PrecomputedSkeleton([(0, 0, 0), (1, 0, 0)], [(0, 1)])
assert PrecomputedSkeleton.equivalent(identity, identity)
diffvertex = PrecomputedSkeleton([(0, 0, 0), (0, 1, 0)], [(0, 1)])
assert not PrecomputedSkeleton.equivalent(identity, diffvertex)
single1 = PrecomputedSkeleton([(0, 0, 0), (1, 0, 0)], edges=[(1, 0)])
single2 = PrecomputedSkeleton([(0, 0, 0), (1, 0, 0)], edges=[(0, 1)])
assert PrecomputedSkeleton.equivalent(single1, single2)
double1 = PrecomputedSkeleton([(0, 0, 0), (1, 0, 0)], edges=[(1, 0)])
double2 = PrecomputedSkeleton([(0, 0, 0), (1, 0, 0)], edges=[(0, 1)])
assert PrecomputedSkeleton.equivalent(double1, double2)
double1 = PrecomputedSkeleton([(0, 0, 0), (1, 0, 0), (1, 1, 0)],
edges=[(1, 0), (1, 2)])
double2 = PrecomputedSkeleton([(0, 0, 0), (1, 0, 0), (1, 1, 0)],
edges=[(2, 1), (0, 1)])
assert PrecomputedSkeleton.equivalent(double1, double2)
double1 = PrecomputedSkeleton([(0, 0, 0), (1, 0, 0), (1, 1, 0), (1, 1, 3)],
edges=[(1, 0), (1, 2), (1, 3)])
double2 = PrecomputedSkeleton([(0, 0, 0), (1, 0, 0), (1, 1, 0), (1, 1, 3)],
edges=[(3, 1), (2, 1), (0, 1)])
assert PrecomputedSkeleton.equivalent(double1, double2)
def trace(
labels,
DBF,
scale=10,
const=10,
anisotropy=(1, 1, 1),
soma_detection_threshold=1100,
soma_acceptance_threshold=4000,
pdrf_scale=5000,
pdrf_exponent=16,
soma_invalidation_scale=0.5,
soma_invalidation_const=0,
fix_branching=True,
):
"""
Given the euclidean distance transform of a label ("Distance to Boundary Function"),
convert it into a skeleton using an algorithm based on TEASAR.
DBF: Result of the euclidean distance transform. Must represent a single label,
assumed to be expressed in chosen physical units (i.e. nm)
scale: during the "rolling ball" invalidation phase, multiply the DBF value by this.
const: during the "rolling ball" invalidation phase, this is the minimum radius in chosen physical units (i.e. nm).
anisotropy: (x,y,z) conversion factor for voxels to chosen physical units (i.e. nm)
soma_detection_threshold: if object has a DBF value larger than this,
root will be placed at largest DBF value and special one time invalidation
will be run over that root location (see soma_invalidation scale)
expressed in chosen physical units (i.e. nm)
pdrf_scale: scale factor in front of dbf, used to weight dbf over euclidean distance (higher to pay more attention to dbf) (default 5000)
pdrf_exponent: exponent in dbf formula on distance from edge, faster if factor of 2 (default 16)
soma_invalidation_scale: the 'scale' factor used in the one time soma root invalidation (default .5)
soma_invalidation_const: the 'const' factor used in the one time soma root invalidation (default 0)
(units in chosen physical units (i.e. nm))
fix_branching: When enabled, zero out the graph edge weights traversed by
of previously found paths. This causes branch points to occur closer to
the actual path divergence. However, there is a large performance penalty
associated with this as dijkstra's algorithm is computed once per a path
rather than once per a skeleton.
Based on the algorithm by:
M. Sato, I. Bitter, M. Bender, A. Kaufman, and M. Nakajima.
"TEASAR: tree-structure extraction algorithm for accurate and robust skeletons"
Proc. the Eighth Pacific Conference on Computer Graphics and Applications. Oct. 2000.
doi:10.1109/PCCGA.2000.883951 (https://ieeexplore.ieee.org/document/883951/)
Returns: Skeleton object
"""
dbf_max = np.max(DBF)
labels = np.asfortranarray(labels)
DBF = np.asfortranarray(DBF)
soma_mode = False
# > 5000 nm, gonna be a soma or blood vessel
# For somata: specially handle the root by
# placing it at the approximate center of the soma
if dbf_max > soma_detection_threshold:
del DBF
labels = ndimage.binary_fill_holes(labels)
labels = np.asfortranarray(labels)
DBF = edt.edt(labels, anisotropy=anisotropy, order='F')
dbf_max = np.max(DBF)
soma_mode = dbf_max > soma_acceptance_threshold
if soma_mode:
root = np.unravel_index(np.argmax(DBF), DBF.shape)
soma_radius = dbf_max * soma_invalidation_scale + soma_invalidation_const
else:
root = find_root(labels, anisotropy)
soma_radius = 0.0
if root is None:
return PrecomputedSkeleton()
# DBF: Distance to Boundary Field
# DAF: Distance from any voxel Field (distance from root field)
# PDRF: Penalized Distance from Root Field
DBF = kimimaro.skeletontricks.zero2inf(DBF) # DBF[ DBF == 0 ] = np.inf
DAF = dijkstra3d.euclidean_distance_field(labels,
root,
anisotropy=anisotropy)
DAF = kimimaro.skeletontricks.inf2zero(DAF) # DAF[ DAF == np.inf ] = 0
PDRF = compute_pdrf(dbf_max, pdrf_scale, pdrf_exponent, DBF, DAF)
# Use dijkstra propogation w/o a target to generate a field of
# pointers from each voxel to its parent. Then we can rapidly
# compute multiple paths by simply hopping pointers using path_from_parents
if not fix_branching:
parents = dijkstra3d.parental_field(PDRF, root)
del PDRF
else:
parents = PDRF
if soma_mode:
invalidated, labels = kimimaro.skeletontricks.roll_invalidation_ball(
labels,
DBF,
np.array([root], dtype=np.uint32),
scale=soma_invalidation_scale,
const=soma_invalidation_const,
anisotropy=anisotropy)
paths = compute_paths(root, labels, DBF, DAF, parents, scale, const,
anisotropy, soma_mode, soma_radius, fix_branching)
skel = PrecomputedSkeleton.simple_merge(
[PrecomputedSkeleton.from_path(path) for path in paths]).consolidate()
verts = skel.vertices.flatten().astype(np.uint32)
skel.radii = DBF[verts[::3], verts[1::3], verts[2::3]]
return skel
def test_downsample():
skel = PrecomputedSkeleton(
[(0, 0, 0), (1, 0, 0), (1, 1, 0), (1, 1, 3), (2, 1, 3), (2, 2, 3)],
edges=[(1, 0), (1, 2), (2, 3), (3, 4), (5, 4)],
radii=[1, 2, 3, 4, 5, 6],
vertex_types=[1, 2, 3, 4, 5, 6],
segid=1337,
)
def should_error(x):
try:
skel.downsample(x)
assert False
except ValueError:
pass
should_error(-1)
should_error(0)
should_error(.5)
should_error(2.00000000000001)
dskel = skel.downsample(1)
assert PrecomputedSkeleton.equivalent(dskel, skel)
assert dskel.id == skel.id
assert dskel.id == 1337
dskel = skel.downsample(2)
dskel_gt = PrecomputedSkeleton([(0, 0, 0), (1, 1, 0), (2, 1, 3),
(2, 2, 3)],
edges=[(1, 0), (1, 2), (2, 3)],
radii=[1, 3, 5, 6],
vertex_types=[1, 3, 5, 6])
assert PrecomputedSkeleton.equivalent(dskel, dskel_gt)
dskel = skel.downsample(3)
dskel_gt = PrecomputedSkeleton(
[(0, 0, 0), (1, 1, 3), (2, 2, 3)],
edges=[(1, 0), (1, 2)],
radii=[1, 4, 6],
vertex_types=[1, 4, 6],
)
assert PrecomputedSkeleton.equivalent(dskel, dskel_gt)
skel = PrecomputedSkeleton([(0, 0, 0), (1, 0, 0), (1, 1, 0), (1, 1, 3),
(2, 1, 3), (2, 2, 3)],
edges=[(1, 0), (1, 2), (3, 4), (5, 4)],
radii=[1, 2, 3, 4, 5, 6],
vertex_types=[1, 2, 3, 4, 5, 6])
dskel = skel.downsample(2)
dskel_gt = PrecomputedSkeleton([(0, 0, 0), (1, 1, 0), (1, 1, 3),
(2, 2, 3)],
edges=[(1, 0), (2, 3)],
radii=[1, 3, 4, 6],
vertex_types=[1, 3, 4, 6])
assert PrecomputedSkeleton.equivalent(dskel, dskel_gt)
